Main

PPIs are critical in the regulation of cellular functions, and their modulation through small molecule inhibition would enable better understanding of key biological events and lead to the development of new molecular therapeutics1,2. Unfortunately, these binding interfaces do not usually contain well-defined, deep binding pockets, making them difficult to target with small molecule inhibitors3,4. To overcome this lack of concavity, molecules are often increased in size in an attempt to improve their affinity for the desired target. Whereas this approach is often acceptable for biological probe molecules5, it represents a real problem for drug development. A major challenge lies in balancing the potency and selectivity of an inhibitor with increased molecular mass, which can jeopardize desired physical properties6.

Recently, targeted covalent inhibition (TCI) has emerged as a potentially more efficient strategy to block PPIs7,8,9. This approach uses bifunctional molecules that consist of a specificity group and a functional warhead capable of generating a covalent linkage with the protein. The former serves as a template by binding to the target protein, thereby allowing the latter to undergo covalent modification. As there is generally an upper limit to the binding affinity that can be achieved for noncovalent ligands10,11,12 and protein-ligand complexes that rely on covalent interactions dramatically favor bond formation13, the TCI approach can enable generation of high-affinity ligands of much lower molecular mass, and also provide other advantages, such as longer residence time and sustained inhibition9,14. Covalent inhibitors that target nonconserved protein residues may provide substantial selectivity over related protein family members or variant forms of the same protein. This is of particular note in the area of oncology, where key cellular proteins are often known to gain resistance to existing therapeutics via point mutations15,16,17. Ligands that contain covalent warheads can potentially aid in overcoming the problem of resistance, while also increasing the therapeutic window18.

Covalent warheads must not be merely highly reactive, however, because this could lead to indiscriminate protein modification and toxicity19. This problem can be mitigated through the use of weak electrophiles, such as electron-deficient olefins that covalently modify cysteines in a reversible, covalent manner20,21. Unfortunately, PPI targets do not always contain free cysteine residues in their protein binding groove, and therefore we wanted to investigate warheads capable of covalently modifying other noncatalytic residues. Of particular interest was the ɛ-amino group of lysine, which is normally overlooked as a site of covalent modification owing to its protonated state under physiological conditions, except in rare cases22. An approach that can be used to modify solvent exposed lysines would truly expand the range of proteins that can be targeted for covalent inhibition, and thus gain the advantages described above23.

During our investigations, we were inspired by a report of aryl boronic acid carbonyls being used to covalently modify multiple lysine residues on the surface of proteins24. We hypothesized that incorporating the same warhead into a small molecule known to bind a desired target would allow site-specific reversible covalent modification of a lysine residue. Moreover, the low reactivity and reversible nature of the warheads would prevent random covalent modification of undesired proteins25. The impact of this approach could allow the generation of small molecule inhibitors capable of modulating PPI pathways without the need to grow ligands to unproductive proportions, and thereby provide a new method for inhibiting these difficult targets26.

In the area of PPI inhibition, Myeloid cell leukemia 1 (Mcl-1) stands out as a key survival factor in a wide range of human cancers, because it is among the top ten most frequently amplified genes in tumors27,28,29,30. Moreover, its high expression levels have been linked to the pathogenesis of a variety of refractory cancers, including multiple myeloma and acute myeloid leukemia as well as poor prognosis in breast cancer31,32,33,34. Mcl-1 functions as an anti-apoptotic protein that neutralizes pro-apoptotic proteins Bak and Bax at the mitochondrial outer membrane, thereby preventing activation of intrinsic apoptosis. In addition, Mcl-1 binds and neutralizes BH3-only proteins, for example, Bim, Noxa and Puma, preventing the activation of pro-apoptotic proteins28,35. Tumor cells often exploit this physiological survival mechanism by overexpressing Mcl-1 to gain continuous resistance to apoptosis. Given this clear therapeutic potential, many medicinal chemistry programs have attempted to develop Mcl-1 inhibitors. Approaches have included antisense oligonucleotides36, stapled peptides37,38 and small molecules, with the latter being identified through traditional39 and fragment-based screening40. Unfortunately, thus far clinical proof of concept has not been achieved41. For these reasons, we were intrigued to investigate whether a boronic acid carbonyl warhead could be used to generate reversible covalent inhibitors of Mcl-1. Here we report our proof-of-concept study, using aryl boronic acid carbonyls as warheads for the reversible covalent targeting of the ɛ-amino side chain of Lys234 in Mcl-1.

Results

Reversible binding to lysine side chain at pH 7.5

To investigate the reversible nature of the iminoboronate bond, we dissolved either 2-formylphenyl boronic acid, 1 or 2-acetylphenyl boronic acid, 2, and (S)-2 acetamido-6-amino-N′-methyl hexanamide, 3, in D2O, and performed a serial dilution with D2O (Online Methods). We analyzed all samples by 1H NMR spectroscopy, which revealed that at high concentrations the equilibrium favored the formation of the iminoboronate and that this bond cleaved upon dilution (Supplementary Results, Supplementary Fig. 1).

Rational design of reversible covalent Mcl-1 inhibitors

Next, we attempted to design a reversible covalent inhibitor by incorporating a boronic acid carbonyl warhead into a previously reported indole-acid-based Mcl-1 inhibitor (4; Fig. 1a)40,42.

Figure 1: Structure-based design of new covalent Mcl-1 inhibitors targeting Lys234.
figure 1

(a) Structure of 4. (b) Docked structure of inhibitor 5 (yellow), illustrating the feasibility of targeting Lys234 using an aryl boronic acid carbonyl warhead.

Full size image

Using a published X-ray crystal structure of Mcl-1 (Protein Data Bank (PDB): 3WIX)39, we determined that the accessibility and proximity of Lys234 offered the best opportunity for reversible covalent modification. This residue, however, was highly flexible in crystal structures, and therefore we used traditional noncovalent docking as covalent docking methods rely on the exact placement of Lys234.

Building from the 7 position of indole 4 in silico, we incorporated several groups capable of placing the warhead proximal to the lysine side chain43. We subsequently ranked the combinations, and selected those that simultaneously placed the warhead carbonyl within ∼3 Å of Lys234 while maintaining the key binding features of the original ligand (Online Methods).

Our models predicted that incorporation of the warheads into 4 via a pyrazole ring and using a phenol linker43 would ideally situate them in the vicinity of Lys234 (Fig. 1b). We therefore synthesized aldehyde boronic acid 5 and ketone boronic acid 6, as well as matched pairs 5,6,7,8,9,10,11 (Table 1 and Supplementary Note).

Table 1 IC50 and EC50 values for inhibitors 5−12
Full size table

Biochemical and cellular potency of Mcl-1 inhibitors

To investigate the effect of those warheads on potency, we obtained measurement of half-maximal inhibitory concentration (IC50) for analogs 5,6,7,8,9,10,11 using a time-resolved fluorescence resonance energy transfer (TR-FRET)-based Mcl-1 binding assay (Table 1 and Online Methods).

From these data, we clearly observed an increase in potency when we incorporated the warheads into the indole acid scaffold. A comparison between parent molecule 10 (ref. 43), with an unsubstituted phenyl group, and analogs with only an aldehyde (7), a ketone (8) or a boronic acid (9), demonstrated that the constituent parts of the warheads indeed conferred greater potency toward Mcl-1. We did, however, observe a further increase in potency upon the incorporation of both groups into a single molecule, as seen with 5 and 6 (Table 1).

Having demonstrated that the warheads increased the potency of the ligands toward Mcl-1 in a biochemical format, we switched our focus to examine their cellular potency. We accomplished this using an assay that measures the activity of caspase-3 and caspase-7 (caspase-3/7) in an Mcl-1-dependent multiple myeloma cell line, MOLP-8 (ref. 44). In agreement with our previous biochemical data, we observed that the warheads contributed to a marked improvement in caspase activation (Table 1). As an example, we detected >24-fold enhancement in half-maximal effective concentration (EC50) between noncovalent inhibitor 10 and warhead containing 5. Moreover, alkylation of the indole nitrogen of this scaffold has been demonstrated previously to increase potency43, and this observation translated to our compounds, with 11 displaying a sixfold increase in cellular potency over the nonmethylated version 5 (Table 1).

Cellular targets of warhead-containing compounds

Having demonstrated that compounds with an incorporated aryl boronic acid carbonyl warhead showed activity against a Mcl-1-dependent cell line, we were keen to demonstrate that the observed activation of apoptosis was due to inhibition of the Mcl-1 pathway, as opposed to nonspecific cytotoxicity associated with the warheads. To address this concern, we first evaluated the ability of 5, 11, and 12 to induce apoptosis in a panel of multiple myeloma cell lines comprised of validated Mcl-1-dependent and -independent cells45. We observed a marked difference in potency for both compounds 5 and 11 across the cell panel, demonstrating activity against those cell lines known to be Mcl-1-dependent (L-363, LP-1, NCI-H929 and the aforementioned MOLP-8), while proving inactive against non-Mcl-1-dependent cell lines (KMS-12-PE and MM.1S) (Fig. 2 and Supplementary Table 3). Although 12 displayed a similar selectivity pattern, the effect was far less dramatic owing to its lower potency across the cell panel (Supplementary Table 3).

Figure 2: Covalent compounds selectively induced apoptosis in Mcl-1-dependent cell lines.
figure 2

(a,b) Caspase-3/7 activity measured after a 2-h incubation with compounds 5 (a) and 11 (b) in Mcl-1-dependent multiple myeloma cells lines (blue), and in Mcl-1 independent cell lines (black). Experiments were run as duplicates.

Full size image

To further validate our hypothesis, we investigated whether the cellular activity of compound 11 could be attenuated in Bak-deficient cells. Bak and Bax are the major pro-apoptotic effector proteins, and are known to be required for the activation of the intrinsic apoptosis pathway by BH3 mimetics27,28. To this end, we treated the known Mcl-1-dependent cell line NCI-H23 (ref. 46) with short interfering (si)RNA targeted to Bak mRNA for 72 h, with immunoblot analysis confirming depletion of Bak (Fig. 3a). Then we incubated inhibitor 11 with the cells, and we observed a marked decrease in its cellular activity when compared to NCI-H23 cells treated with a random siRNA (Fig. 3b). Moreover, when we incubated either NCI-H23 or NCI-H23 Bak-deficient cells with cytotoxic agent staurosporine, we observed no such erosion in activity (Fig. 3c).

Figure 3: Activity of compound 11, but not staurosporine, required Bak.
figure 3

(a) Knockdown of Bak protein in NCI-H23 cells assessed by immunoblot after 72 h. Full gel images are in Supplementary Figure 2. (b,c) Caspase-3/7 activity in NCI-H23 cells treated with siRNA targeting Bak mRNA for 72 h before treatment with compound 11 (b) or staurosporine (c) for 6 h. Experiments were run as duplicates.

Full size image

Finally, with evidence in hand that our compounds were killing cells in an Mcl-1-dependent manner, we investigated whether the observed activity was due to a direct disruption of the Mcl-1−Bak complex. Briefly, we treated MOLP-8 cells with compound 5 for 30 min and detected Mcl-1 complexes by immunoprecipation (IP) and immunoblot. Mcl-1 was dissociated from Bak at 1 μM and 10 μM (Fig. 4a), which is consistent with the observed induction of caspase-3/7 at these concentrations for compound 5. Moreover, there were no changes in total protein levels of Mcl-1 or Bak at these concentrations (Fig. 4b), suggesting the observed decrease in Mcl-1−Bak complexes was not a result of downregulation or turnover of Mcl-1 from a nonspecific cytotoxic effect.

Figure 4: Compound 5 disrupted Mcl-1 complexes in MOLP-8 cells at concentrations that induced apoptosis.
figure 4

(a) Immunoprecipitation (IP) of Mcl-1 from lysates of MOLP-8 cells treated with tenfold serial dilutions of compound 5 for 30 min, followed by immunoblot analysis. (b) Analysis of total protein amounts for Mcl-1 or Bak in lysates of cells treated as in a. Full gel images are in Supplementary Figure 3.

Full size image

Determination of covalent modification of Mcl-1 by mass spectrometry

Next, we wanted to confirm covalent modification of Mcl-1 via liquid chromatography-mass spectroscopy (LC-MS). Initially, we attempted to trap the covalent adduct via sodium cyanoborohydride reduction (Supplementary Fig. 4). This proved unnecessary, and LC-MS analysis of Mcl-1 treated with inhibitor 5 revealed a peak in the LC trace with a mass of 18,799 Da, corresponding to addition of one molecule of 5 and the loss of water. We also observed a mass corresponding to the loss of a second molecule of water (18,781 Da) (Supplementary Fig. 5). This result supports the hypothesis that these warheads can covalently modify Mcl-1 through an as-yet unidentified lysine side chain. In contrast, we observed no covalent adduct when we incubated Mcl-1 with 2-formyl-4-methoxyphenyl boronic acid (13), which lacked the noncovalent binding scaffold, aldehyde (7) or boronic acid (9) matched pairs (Supplementary Fig. 5).

Finally, to gain additional confidence in the covalent linkages, we monitored the rate of adduct formation between 5 or 6 and Mcl-1 by LC-MS. We observed a much faster rate of reaction with the aldehyde containing 5, yielding ∼50% conversion within 1 h (Supplementary Fig. 6a), whereas ketone 6 required almost 8 h to reach the same 50% conversion (Supplementary Fig. 6b). We observed no further increase in conversion at later time points, presumably because the reaction reached equilibrium. In similar LC-MS experiments with varying concentrations of 5, the rate of adduct formation was also concentration-dependent (Supplementary Fig. 7).

Binding kinetics of Mcl-1 inhibition

Having demonstrated that boronic acid carbonyl warheads can generate covalent adducts with Mcl-1 under physiological conditions, we next focused on the reversibility of the modification. To examine this, we used surface plasmon resonance (SPR) to study the kinetics of binding for the compounds against immobilized Mcl-1 (Supplementary Table 4). The results from these biophysical studies confirmed enhanced affinity upon incorporation of the warheads despite a short contact time, which was consistent with previously determined IC50 measurements. The binding dissociation for 5 and 6 displayed a biphasic profile and did not fit well to a 1:1 binding model. In contrast, when we fitted the data using a two-state binding model, a second slower dissociation rate became apparent (kd2 for 5 was 0.018 1/s, and kd2 for 6 was 0.010 1/s; Supplementary Fig. 8).

To further study the binding kinetics of these compounds, we examined the time-dependent inhibition of Mcl-1. We performed binding experiments in solution and determined IC50 at multiple time points, using a modified procedure of our TR-FRET assay (Online Methods). For inhibitors 5 and 6, we observed a decrease in fluorescence signal over time as compounds competed with the BIM peptide for binding to Mcl-1, thus confirming a time-dependent occupancy of Mcl-1 and suggesting a covalent binding mode (Supplementary Fig. 9a,b). In contrast, no time-dependent inhibition was displayed by ketone analog 8 or Mcl-1 inhibitor 12, which we chose as a comparator because of its similar potency to 5 and 6 in our biochemical assay (Supplementary Fig. 9c,d).

Identification of K234 as the site of modification

Although compounds 5 and 6 could reversibly covalently modify Mcl-1, the site of modification remained unknown. From our molecular modeling studies we predicted that Lys234 was the most likely residue undergoing modification. We performed site-directed mutagenesis to replace Lys234 with an alanine to investigate the effect of substitution on both covalent bond formation and potency. For this, we recombinantly expressed the Mcl-1 K234A variant and purified it. We subsequently demonstrated that this variant protein behaved similarly to wild-type Mcl-1 with reference compound 12, yielding a similar IC50 in our TR-FRET binding assay (Supplementary Fig. 10). Next, we examined the biochemical potency of 6 and its non-warhead-containing matched pair 8 against the Mcl-1 K234A variant. As expected, noncovalent compound 8 possessed a similar IC50 value when compared to wild-type Mcl-1 (Fig. 5a). In contrast, inhibitor 6 displayed a marked reduction in potency with Mcl-1 K234A (Fig. 5b). Moreover, mass spectrometry analysis of Mcl-1 K234A treated with either 5 or 6 did not show any covalent complex formation (Supplementary Fig. 11).

Figure 5: Effect of Lys234 mutation on the biochemical potency.
figure 5

(a,b) Inhibition of wild-type Mcl-1 and Mcl-1 Lys234 by compound 8 (a) and 6 (b). IC50 measurements were obtained by TR-FRET-based binding assay. Experiments were run as duplicates.

Full size image

Discussion

Over the past decade there has been a renewed interest in the area of covalent inhibition, particularly in the field of oncology8, and new therapeutics that use a covalent linkage to site-specifically target noncatalytic residues are now entering the market18. Nonetheless, the ability to target a desired protein for covalent inhibition under physiological conditions is often limited to reaction with cysteine residues. Recently, it has been reported that boronic acid carbonyl warheads can enable rapid and reversible modification of the ɛ-amino group of lysine residues24,47. Here we demonstrated that this technology can be used site-specifically to expand the scope of residues that can be targeted in a reversible covalent manner21. Although our investigations focused on the inhibition of a PPI target, Mcl-1, this approach should allow for the generation of reversible covalent inhibitors against other protein classes.

Our investigation of the reversible nature of the iminoboronate bond under physiological pH using 1H NMR spectroscopy demonstrated that at high concentrations of the warhead and lysine, the equilibrium favored the formation of the iminoboronate whereas hydrolysis occurred upon dilution (Supplementary Fig. 1)24. The reversible nature of the bond suggests that inhibitors that use the boronic acid carbonyl warhead to generate a covalent linkage will benefit from the advantages of reversible covalent inhibitors48,49. In addition, the fact that imine formation is first-order in both the warhead and amine suggests that targeting a lysine in a well-defined binding pocket may lead to more substantial increases in potency and slower off rates, when compared to the reversible covalent inhibitors reported here. This technology may therefore find applications in the generation of covalent kinase inhibitors, an area that has already proven its therapeutic potential18. The solvent-sequestered ATP binding site of a kinase would provide the ideal conditions for formation of the imine bond. Moreover, the ability to target lysines would expand the number of kinases that can be targeted for covalent inhibition, which in the past have relied on the fortuitous placement of a cysteine side chain.

In our study, we incorporated the warhead into a previously disclosed noncovalent Mcl-1 inhibitor42. From our molecular modeling and crystal structure analysis, we determined that incorporating a pyrazole aryl ring and a phenolic linker in to 4 would situate the warhead proximal to Lys234 and facilitate generation of the desired iminoboronate (Fig. 1). From biochemical and cellular examination of the synthesized compounds, we observed an increase in potency upon incorporation of the warhead (Table 1). Although we could not attribute the greater binding affinity to covalent modification, we were pleased to note that the cellular activity demonstrated that the warheads did not appear to affect the compounds' ability to traverse the cell membrane. In addition, we also observed that previously described SAR translated to the warhead-containing compounds, i.e., N-alkylation of the indole increased potency for Mcl-1 (Table 1)43.

To provide evidence that the warhead containing ligands formed covalent bonds with Mcl-1, we used SPR to determine binding kinetics and confirmed our findings through time-dependent inhibition measurements in a biochemical format. These two sets of data demonstrated that the compounds operated in a covalent manner, and the SPR study revealed the reversible nature to the bond and demonstrated a prolonged duration of Mcl-1 inhibition compared to noncovalent inhibitors. To provide orthogonal evidence of covalent bond formation, we investigated the mass spectrum of Mcl-1 treated with compounds 5 and 6. In both cases we observed a mass-to-charge ratio that related to the formation of a single covalent adduct when we incubated Mcl-1 with either compound. Initially, we attempted to remove the reversible nature of the warhead via a sodium cyanoborohydride reduction of the imine (Supplementary Fig. 4). To our surprise, we discovered this was unnecessary, because we observed the covalent adduct between the inhibitor and Mcl-1 without reduction, despite the reversible nature of the bond. Although we cannot be certain as to the reason for this unexpected result, it is likely due to the lack of complete unfolding of the protein under LC-MS conditions. For reversible covalent inhibition, this has been observed previously50. More importantly, the observation of the adduct provides compelling evidence for covalent bond formation, which is further attested to by the fact that its formation was both time- and concentration-dependent. In addition, the constituent parts, 7 and 8, and the warhead alone, 13, did not generate covalent adducts when incubated with Mcl-1. This suggests that both elements of the warhead and the noncovalent recognition group are needed for covalent bond formation, in line with our original hypothesis.

As a whole, these data describe the discovery of, to our knowledge, the first reversible covalent inhibitors of a PPI target, Mcl-1, and demonstrate that a boronic acid carbonyl warhead can be rationally designed into small molecules to generate ligands capable of forming a reversible covalent bond with a desired lysine side chain. In our hands, this led to biochemical and cellular potency increases, while also decreasing the off rate of the ligand. These reversible covalent inhibitors elicited caspase activation in a Mcl-1-dependent manner, and were not acting as general cytotoxics. This demonstrates that for these molecules the warhead portion was tolerated by cells.

In the future, targeted reversible covalent modification of noncatalytic lysines can be broadly applied to other protein families to generate highly selective reversible inhibitors with improved residence times and minimal off-target effects. In addition, because of the residue specificity of this technology, it may allow the targeting of oncogenic lysine substitutions and provide selectivity versus the wild-type protein. Ultimately, this methodology will find a variety of applications in fields ranging from chemical biology, where covalent probes can be constructed to target lysine residues for pull-downs, to drug discovery, where it can be used to inhibit high-value targets, such as Mcl-1, with more drug-like molecules.

Methods

Chemistry.

Synthetic schemes, detailed procedures and characterization of compounds 5,6,7,8,9,10,11 can be found in the Supplementary Note.

Computational details for inhibitor modeling.

The model for 5 bound to Mcl-1 (Fig. 1b) was generated as follows. The crystal structure of Mcl-1 with 7-(4-carboxyphenyl)-3-(3-(naphthalen-1-yloxy)propyl)-pyrazolo[1,5-a]pyridine-2-carboxylic acid from PDB 3WIX39 was prepared for docking using the Schrodinger Suite of software which includes Maestro, Prime, LigPrep and Glide (Protein Preparation Wizard; Epik version 3.2; Impact version 6.7; Prime version 4.0; Glide, version 6.7, Schrödinger, LLC, 2015). First the Protein Preparation workflow within Maestro was used, and His224 was protonated on Nδ, missing side chains were built with Prime, sequence termini were capped, hydrogen atoms minimized, and crystallographic water molecules were deleted. Independently, ligand tautomer and protonation states were prepared from SMILES strings using Leatherface51 and initial conformations were generated with Corina52. These were refined using LigPrep, where up to 32 stereoisomers and four ring conformations could be generated. Any ring conformations with energy >5 kcal/mol were removed. Minimization was then carried out with the OPLS 2005 forcefield. The ligands were then docked to the protein using Glide SP 5.0 using settings (FORCE PLANAR = 2 and POSTDOCK = FALSE), retaining five poses per ligand. Finally, these poses were manually reviewed for those that placed the warhead C=O near the targeted Lys234 and maintained a linker conformation that was not highly strained. The Lys234-5 interaction was further refined using Prime, first constraining the distance from a Lys234 HZ atom to the ligand carbonyl oxygen to 2.5 Å, then allowing the ligand and Lys234 sidechain to minimize further within the pocket to yield an optimized pose. Figure 1 and the graphical abstract were prepared with MOE (Chemical Computing Group Inc. Molecular Operating Environment (MOE) (2014.09)).

Mass spectrometric analysis of covalent modification.

LC-MS was performed on a Waters Acquity UPLC system coupled to a Waters LCT Premier mass spectrometer equipped with an electrospray probe operated in positive ionization mode. 10 μL samples were injected on an Agilent Poroshell 300SB-C8, 2.1 × 75 mm, 5 μM column. The mobile phase was 0.1% (vol/vol) formic acid in water (A) and 0.1% (vol/vol) formic acid in acetonitrile (B). The separation was performed by a 7 min linear gradient from 20% to 72.5% B, then ramped to 95% B over 0.1 min, then held at 95% B for 1 min, all at a flow rate of 0.6 mL/min. Mass spectrometer source conditions were capillary voltage, 3,500 V; cone voltage, 160 V; source temperature, 120 °C; scan range, 100−1,860 a.m.u. with a cycle time of 0.19 s. DAD detection was 210 nm to 400 nm. Protein deconvolution was performed using Waters MaxEnt.

Mutagenesis.

Using pGEX6P Mcl-1 (E171-G327) plasmid as a template, the codon encoding Lys234 was exchanged to a codon encoding alanine using QuickChange kit (Agilent Technologies) following the manufacturer's guidelines. Briefly, pGEX6P Mcl-1 (E171-G327) plasmid was amplified by PCR using the primers, 5′-CAA GGC ATG CTT CGG GCA CTG GAC ATC AAA AAC-3′ (forward) and 5′-GTT TTT GAT GTC CAG TGC CCG AAG CAT GCC TTG-3′ (reverse) corresponding to K234A mutation, followed by DpnI (restriction enzyme from Escherichia coli strain carrying Diplococcus pneumoniae gene) digestion and transformation into TOP10 Competent Cells (Invitrogen, Life Technologies). A single colony of intended sequence bearing the mutation encoding the K234A substituted protein was grown overnight in 5 mL Lysogeny Broth (LB) culture at 37 °C and pelleted, DNA was extracted and purified. Successful mutagenesis was confirmed by DNA sequencing.

Protein expression and purification.

Amino-terminal GST-tagged fusions linked with PreScission Protease proteolytic site of wild-type and mutated Mcl-1(E171-G327) plasmids were transformed in Escherichia coli BL2(DE3) strain. Overnight cultures were diluted in 1:100 LB medium supplemented with 100 μg/mL of ampicilin and grown at 37 °C. At OD600 of ∼0.6−0.8, expression was induced by addition of 0.1 mM of isopropyl-β-D-thiogalactopyranoside (IPTG) and cultures were grown at 18 °C overnight. The fusion proteins were soluble and expressed in good yields, and expression was verified by Coomassie blue staining. After cell harvest and French press lysis at 4 °C (25 mM Tris-HCl, 150 mM NaCl, 2 mM DTT, pH 7.4, Roche Complete Protease inhibitor tablet), the solution was clarified by centrifugation at 14,000 r.p.m. (30 min, 4 °C) and cell debris was removed. Supernatant was nutated with Glutathione Sepharose 4B (GE Healthcare Life Sciences) for 1 h at 4 °C, followed by centrifugation at 4,000 r.p.m. for 10 min. Supernatant was removed and the resulting resin was washed three times with the lysis buffer. Resin-bound fusion protein was eluted with elution buffer (100 mM Tris-HCl, 150 mM NaCl, 2 mM DTT, 20 mM reduced glutathione, pH 8.0). A portion of eluted fractions were pooled and buffer exchanged in SnakeSkin dialysis tubing (10 kDa molecular weight cutoff, Life Technologies) against 25 mM Tris-HCl, 150 mM NaCl, 1 mM DTT, pH 7.4 overnight at 4 °C. GST-Mcl-1 and GST-Mcl-1 K234A (E171-G327) fusion proteins were > 95% pure as determined by Coomassie-stained SDS-PAGE and mass spectrometry. After this step, the above fractions were concentrated to ∼2 mg/mL using an Amicon concentrator (10 kDa cutoff, EMD Millipore). A portion of eluted fractions were pooled and dialyzed against PreScission Protease cleavage buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM DTT, pH 7.4) in 10 kDa molecular weight cutoff in SnakeSkin dialysis tubing overnight at 4 °C. GST-cleaved supernatant was further purified by gel-filtration chromatography with a Sephacryl S-200 column (GE Healthcare Life Sciences) in 25 mM Tris-HCl, 150 mM NaCl, 1 mM DTT, pH 7.4. Fractions from the purification were analyzed by Coomassie-stained SDS-PAGE and mass spectrometry. Pure fractions were pooled and concentrated to ∼2 mg/mL using an Amicon concentrator (10 kDa cutoff, EMD Millipore). Untagged human Mcl-1 and Mcl-1 K234A (E171−G327) were used for mass spectrometry experiments. GST-tagged human Mcl-1 was used for biochemical inhibition assays. 6His-tagged human Mcl-1 (E171−G327) was used for surface plasmon resonance (SPR) experiments.

Biochemical inhibition assays and Ki calculation.

TR-FRET assay was used to assess the ability of compounds to disrupt the interaction between recombinant human Mcl-1 and a labeled BIM peptide probe.

Specifically, sequence encoding human Mcl-1 enzyme from Mcl-1 (E171-G327) was cloned into an overexpression vector, expressed as an N-terminal GST-tagged fusion protein in E. coli, and subsequently purified via glutathione sepharose affinity and size-exclusion chromatography as described above. Binding of BIM BH3 domain peptide to Mcl-1 protein was evaluated by a TR-FRET assay. The assay was constructed such that europium used as the fluorescence donor was attached to a GST-Mcl-1 fusion protein by the binding of a europium-labeled anti-GST antibody, and FAM-labeled BIM BH3 domain peptide was used as the fluorescence acceptor.

Compound IC50 values were assessed following a 10-point, half-log10 dilution schema starting at 10 μM compound concentration. The assay was performed in 384-well LV plates (Matrix 4365, Thermo Scientific) and run in the presence and absence of the compound of interest. Each well of 12 μL assay mixture contained 10 mM Tris (pH 7.4), 1 mM DTT, 0.005% Tween-20, 150 mM NaCl, 10% DMSO, 3 nM GST-Mcl-1, 1 nM LanthaScreen Eu-tagged GST antibody (#PV5594, Invitrogen, Life Technologies), and 8 nM fluorescently labeled peptide. Reactions were incubated at 24 °C for 60 min before reading on a Tecan F500 spectrofluorometer with excitation at 340 nm and emission at 620 nm and 665 nm. Subsequently, the ratio of fluorescence emission intensity at 665 nm to 620 nm was calculated for each reaction, and the dose response of the ratio to testing compound concentration was fitted to a select fit model that will provide the best fit quality using automatic parameters to derive IC50 values for each testing compound.

The following equation was used to calculate apparent Ki:

where total concentration of the fluorescence-labeled ligand L = 8.0 nM, and dissociation constant of the protein-ligand complex Kd = 3.0 nM.

Surface plasmon resonance (SPR) binding experiments.

A Biacore T200 instrument (GE Healthcare) was used to monitor binding interactions using a direct binding assay format. 6His-tagged Mcl-1 protein (E171-G327) was immobilized using NTA capture coupling at a flow rate of 10 μL/min and using an immobilization running buffer containing 10 mM HEPES, 150 mM NaCl, 1 mM TCEP and 0.05% Tween-20 at 25 °C. Briefly, the sensor surface was activated with a 1 min injection of 0.5mM NiCl2 and a 7 min injection of a mixture of 11.5 mg/mL N-hydroxysuccinimide with 75 mg/mL 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride. Approximately 300 response units of Mcl-1 protein (2 μg/mL in immobilization running buffer) were immobilized using the 'aim for' function in the T200 Control Software (GE Healthcare). Remaining reactive esters were blocked using a 7 min injection of 1M ethanolamine. Reference flow cells were prepared without protein. All binding measurements were performed in 10 mM Tris, pH 7.5, 300 mM NaCl, 1 mM TCEP, 1% DMSO and 0.02% Tween-20 at 25 °C at a flow rate of 30 μL/min. Buffer was primed through the instrument overnight to stabilize the surface before subsequent assay steps. Samples were injected over the protein and reference surfaces for at least 30 s and included at least 180 s dissociation time. Prior to kinetic analysis, solvent calibration and double referencing subtractions were made to eliminate bulk refractive index changes, injection noise and data drift. Affinity and binding kinetic parameters were determined by global fitting to a 1:1 or a two state reaction model within the Biacore T200 Evaluation Software (GE Healthcare).

Cellular assays.

Cell lines used in this study. NCI-H929 and NCI-H23 cells were obtained from ATCC. KMS-12-PE, L-363, MOLP-8, and LP-1 cells were obtained from DSMZ. MM.1S cells were obtained from Northwestern University. All cell lines tested negative for mycoplasma contamination.

Caspase activation in multiple myeloma (MM) cells. Cells were plated at 15,000 cells/well of a 96-well white opaque plates in OPTIMEM (Life Technologies). Cells were treated with compounds at indicated concentrations for 2 h (37 °C, 5% CO2) in OPTIMEM media with a final DMSO concentration of 0.3%. Caspase-3/7 activation was subsequently determined using a Caspase-Glo 3/7 Assay (Promega Corporation) as described in manufacturer's instructions. Dose-reponse curves were plotted and analyzed (including EC50 determination) using GraphPad Prism.

Bak-dependent apoptosis in NCI-H23 cells. NCI-H23 cells were trypsinized and resuspended to 40,000 cells/mL in RPMI-1640 medium supplemented with 10% FBS and 2 mM L-glutamine. In a 15 ml conical tube, 5 mL of OPTIMEM, 15 μL of 20 μM siRNA targeting Bak (Ambion Silencer(R) Select, catalog s1881, Thermo Fisher Scientific) or 20 μM non-targeting siRNA (4390847, Thermo Fisher Scientific), and 50 μL of Lipofectamine RNAiMAX (Thermo Fisher Scientific) were mixed and incubated at room temperature for 20 min. 15 mL of cell suspension was added to each siRNA master mix. 100 μL of this was dispensed into each well of a 96-well black-wall, clear-bottom flat assay plate, and incubated for 72 h (37 °C, 5% CO2). Media was replaced with OPTIMEM, and cells were treated with compounds at indicated concentrations in OPTIMEM media containing 2 μM CellEvent Caspase-3/7 Green Detection Reagent (Thermo Fisher Scientific) with a final DMSO concentration of 0.3%. Cell plates were imaged every hour for 6 h using the Incucyte(R) ZOOM Live-Cell Analysis System (Essen Bioscience).

Immunoprecipitation.

Cells were cultured in RPMI-1640 with 20% FBS and 2 mM L-glutamine in flasks. Cells were collected into 50 mL centrifuge tubes and resuspended in serum-free medium (5 million/mL). 10 mL of cells was plated into a 10 cm dish (one per treatment), and incubated in a tissue culture incubator (37 °C, 5% CO2) for 1−2 h without compound. Compounds were prepared in DMSO and added to cells for 30 min. Media and cells were removed and transferred into a 15 mL conical tube and centrifuged at 1000 rpm for 5 min. After supernatant was removed, 0.5 mL ice-cold lysis buffer (10 mM HEPES, pH 7.0, 150 mM NaCl, 1% CHAPS, 1 mM EDTA, 5 mM MgCl2, Roche Complete Protease inhibitor tabs (1 tab/10 mL, Roche Diagnostics Corporation), Roche PhosSTOP phosphatase inhibitor tabs (1 tab/10 mL, Roche Diagnostics Corporation) was added to each tube. Pellets were resuspended in lysis buffer, and lysate was collected into a microcentrifuge tube on ice and incubated for 20 min with vortexing every 5 min. After incubation, tubes were centrifuged at 14,000 r.p.m. at 4 °C for 15 min, and supernatants were transferred to a fresh microcentrifuge tube. Protein concentration was measured using a bicinchoninic acid assay (BCA, Thermo Scientific). 1.0 mg of fresh lysate in a volume of 500 μL was used for each IP. Samples were pre-cleared for 30 min using rotation at 4 °C with 20 μL of a 50% slurry of Protein A/G magnetic beads (Thermo Scientific). Lysate was incubated with 10 μg of mouse anti-Mcl-1 antibody (MABC43, EMD Millipore) overnight at 4 °C with rotation. Mcl-1 protein was immunoprecipitated by incubating 20 μL of 50% slurry of Protein A/G magnetic beads with lysates for 1 h at 4 °C with rotation. The beads were washed 4 times with 1.5 mL ice-cold 50% lysis buffer and 50% PBS. 50 μL LDS Sample Buffer (Invitrogen, Life Technologies) with 10% Sample Reducing Agent (Invitrogen, Life Technologies) was added to each IP pellet. Original lysates were also prepared for SDS-PAGE by adding LDS buffer and sample reducing agent to the 2.0 mg/mL lysate.

Western blot.

Lysate and IP samples were heated in a 100 °C heat block for 10 min. 10 μL of each IP (and lysate (13 μg)), were run on a 26-well 10% Bis-Tris NuPAGE in MES buffer. Protein was transferred onto nitrocellulose membranes using the Invitrogen iBlot (Invitrogen, Life Technologies) system (program #3, 7:00 run time). Blots were blocked with 5% milk in Tris-buffered saline with 0.05% Tween (TBST) for 1 h at room temperature. Blots were incubated with rocking at 4 °C overnight with the following primary antibodies in 5% milk (in TBST): Mcl-1 (cat# sc-819, SantaCruz Biotechnology), Bak (cat# V9131, Sigma-Aldrich), and vinculin (cat# V9131, Sigma-Aldrich). After incubation, blots were rinsed once and washed three times for minutes with TBST. For Mcl-1, blots were incubated with a 1:200 dilution of Clean-Blot IP Detection Reagent (cat# 21230, Thermo Scientific) for 1 h at room temperature. For all others, blots were incubated with a 1:4,000 dilution of secondary antibody (HRP-conjugated goat anti-rabbit IgG, Fc fragment cat #111-035-046, Jackson ImmunoResearch) for 1 h at room temperature. Blots were rinsed once and washed four times for 5 min in TBST. SuperSignal WestDura (Thermo Scientific) was used as the HRP substrate (1 min incubation). Bands were detected using a GE ImageQuant LAS 4000.

Additional information

Any supplementary information, chemical compound information and source data are available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html. Correspondence and requests for materials should be addressed to N.P.G. or Q.S.